A HowNet-based Feature Selection Method for Chinese Text Representation

Data dimension reduction plays an important role in the field of text representation. An effective dimension reduction method can not only reduce computation complexity, but help to improve the accuracy of text classification. This paper presents a new method of dimension reduction which is based on words semantic similarities. Being different with traditional methods which usually use the statistical information of words, natural language processing knowledge is used in our method which considers semantic information and POS information of feature terms. The experimental results show that our method is effective in dimensionality reduction of text representation and achieves a higher accuracy of text classification. The semantic similarity based method is a suitable method for text representation.